Kernel 2.6.19 throwing up hd read errors

cr9c1 · 12-10-2006, 12:46 AM

I tried compiling linux-2.6.19 today and have since tried several times recompiling it.

Im getting those pesky Seek/Read errors and DMA ends up getting disabled.

I've verified that I have the proper drivers loading in the kernel. I am using a ReiserFS file system and that is also properly compiled into the kerenl. I've tried everything I can come up with but nothing works.

Anyone have any ideas?

erickFis · 12-11-2006, 05:09 AM

Hi cr9c1!

I had this issue just a few time ago,
and I now how boring is to loose dma....

After running several tests with smard, i found (in mail lists), the solution was to exchange my flat cable.

And, believe or not, it just worked!

Now I'm a happy dma man again

kremers78 · 12-11-2006, 08:23 AM

Try enabling dma support in the kernel. That worked for me (same kernel, also reiserfs)

hollywoodb · 12-11-2006, 09:28 AM

Quote:

Originally Posted by erickFis

Hi cr9c1!

I had this issue just a few time ago,
and I now how boring is to loose dma....

After running several tests with smard, i found (in mail lists), the solution was to exchange my flat cable.

And, believe or not, it just worked!

Now I'm a happy dma man again

This is fairly common... the problem with flat IDE cables is that they are subject to breakage. They're relatively stiff yet often they are folded or bent at semi-sharp angles, which causes metal fatigue and can cause the thin wires inside to break or become semi-seperated where the cable is no longer reliable.

I've since converted all my machines to round cables, which cost a little more. To date I've had zero issues with the round cables, of course you still shouldn't fold them or force them to sharp bends.

cr9c1 · 12-17-2006, 08:36 PM

I tried a different IDE cable, that didn't work. DMA support is enabled in the kernel. It for some reason is just throwing up errors.

I can turn DMA on, it lets me do that, but for some reason when the kernel boots it doesn't work for me.

erickFis · 12-18-2006, 04:54 AM

so it's time to post here your errors log:
please post #dmesg

You should try #smartctl for check your hd either.
Please check #smartctl man page or internet manuals for usage, because just now I can't remember how to do the check up.
And, finally, post your #smartctl output here to.

pbhj · 12-18-2006, 04:01 PM

Yeah I got problems too. I've been running this kernel (2.6.16.13) since May IIRC (which I probably don't).

I get a whole slew of errors like:

Code:

hdb: task_in_intr: status=0x59 { DriveReady SeekComplete DataRequest Error }
hdb: task_in_intr: error=0x10 { SectorIdNotFound }, LBAsect=39167615, sector=39102336
ide: failed opcode was: unknown

I tried recompiling several times with different options. I've tried enabling an initrd (initial ram disk) for the reiserfs modules (even though I had reiser support built in). I tried swapping the ide channels (drive A moved from ide0 to ide1 on mainboard) and new cables. Changed fstab so none of partitions on hdb are mounted automatically.

None of this worked.

I'd have thought that with noauto set on all hdb partitions that the drive on /dev/hdb wouldn't even be touched at boot ... but this is clearly not true. Note that hdb was historically a windows drive but has several partitions now - it used to have lilo in the MBR but that's on the other drive now too.

I too enable DMA using "hdparm -d1 /dev/hdb" in my rc.local (again IIRC).

People told me the disk was failing but smart told me otherwise and the last 7 months use appears to confirm it's not hardware failure - or if it is it's an esoteric failure to respond to a particular opcode that doesn't make any difference to disk / fs integrity. I also noticed the other day when running an fsck (which gave no errors) via knoppix that I get the same sort of errors when I boot knoppix.

That's it.

I'm looking forward to finding time for a kernel upgrade to see if it gets fixed. In the mean time I'm just living with it.

HTH. Any progress then please post back here.